Analysts’ forecasts and financial reporting
quality: Evidence from private
subsidiaries
Master Thesis
By
Floris van Halm
2
MSc Accounting Thesis
Analysts’ forecasts and financial reporting
quality: Evidence from private
subsidiaries
University of Groningen Faculty of Economics and Business
Department of Accounting
June 2019
Floris van Halm
Jan van Scorelstraat 41 C
3583CK Utrecht
+31616677265
f.j.d.van.halm@student.rug.nl
Student number: 2758466
First supervisor: S. Rusanescu, PhD
3
Abstract
This study examines the relation between subsidiary financial reporting quality (FRQ) and the forecast accuracy of analysts following the parent of the multinational corporation (MNC). Prior studies suggest that MNCs prefer to manage earnings through their subsidiaries. Moreover, there is evidence that FRQ positively influences the accuracy of analysts’ forecasts. Therefore, the study predicts and finds that subsidiary FRQ is positively associated with the forecast accuracy of analysts following the parent company. Higher subsidiaries FRQ should make it simpler for analysts to make reliable forecasts about the parent because it helps to reduce the information asymmetry between managers and market participants. Additionally, I follow previous research documenting that a country’s institutional quality (IQ) is negatively related to the extent of earnings management practices that firms engage in. Specifically, I investigate the effect of the countries’ IQ where subsidiaries are located on the relationship between the subsidiary FRQ and the analysts’ forecast accuracy. Particularly, I find that the countries’ IQ of the subsidiary of the MNC does not influence the relationship between the subsidiaries’ FRQ and the forecast accuracy of analysts for the parent MNC, which is included in the sample of the study. This study is based on 12,344 forecast observations from 312 U.S. MNCs and their 1,797 European subsidiaries’ FRQ obtained in a two-year time period between 2015 and 2017. In sum, my findings suggest that subsidiaries FRQ is positively associated with forecast accuracy of analysts following the parent MNC and that IQ does not influence the relationship between subsidiaries FRQ and forecast accuracy of analysts about the parent MNC.
Keywords: Financial reporting quality, subsidiaries, multinational firms, analyst’s forecasts,
4
Table of Contents
Introduction 5 Literature 8 Literature review 8 Hypothesis development 11 Methodology 14 Sample selection 14 Forecast accuracy 14Financial reporting quality 15
Institutional quality 16 Empirical model 17 Control variables 17 Results 20 Descriptive statistics 20 Correlations 22
Results regression analysis 26
Discussion and Conclusion 29
5
Introduction
Recently, the OECD (2018) published a report about the functioning of MNCs and their immense impact on the worldwide economy. The OECD emphasizes that MNCs are responsible for nearly 28% of the gross domestic product and even 33% of the global output, while they account for such a small proportion of all the companies in the world. A multinational can be defined as ‘a company that has a country location at the place where the firm is incorporated, thereby they have an establishment of branches or subsidiaries in foreign countries’ (Lazarus, 2001). Multinationals choose to make investments in foreign countries for various reasons. For instance, to reduce the average labor costs, to get better access to natural resources, gain more specific technologic/knowledge and to gather more cost efficient capital (Kinoshita and Campoa, 2003; Lee, 2015; OECD, 2018).
These investments gain more interest by shareholders, due to the opportunity an oversea investment gives to increase the profits of a firm (Lee, 2015). However, the oversea investments also create an opportunity for management activities that are unfavorable for shareholders and even stakeholders. When share- and stakeholders do not have sufficient resources, incentives, or access to relevant information to monitor manager’s actions, earnings management practices become more prevalent (Schipper, 1989). Earnings management (EM) is a technique used by managers in the reporting process with the intention to influence the earnings of the firm in their favor (Schipper, 1989). Managers have incentives to increase earnings, because of binding debt covenants, manager compensation agreement, equity offerings and insider trading (DeAngelo, DeAngelo and Skinner, 1994; Beneish, 2001; Bergstresser and Philippon, 2006). On the other hand, managers are incentivized to decrease earnings temporarily to influence the likelihood of negotiated- or regulatory outcomes or to decrease earnings with the intention to reduce the tax expenses of the firm (Beneish, 2001; Maydew, 1997).
Managers of MNCs have more opportunities to engage in EM practices than those of domestic firms because of the organizational complexity that arises due to the corporate international diversification, leading to more information asymmetry between managers and investors (Chin, Chen and Hsieh, 2009). Besides, analysts and investors have more difficulties to obtain information about foreign investments than domestic investments (Bodnar and Weintrop, 1997). Managers of MNCs make use of that and engage more in EM activities at oversea investments (Durnev, Li and Magnan, 2009; Fan, 2012). The overseas investments are more difficult to control and experience more risks or uncertainty due to differences in regulations, economic growth, political climate and competitive environment (Yang, Hu and Giacomino, 2010; Christophe, 2002). The EM practices of managers of MNCs that exist due to
6
information asymmetry may decrease firm value of the MNC (Chin et al., 2009). Information asymmetry between managers and investors should be reduced to prevent a decrease in firm value (Chin et al., 2009). Previous research states that financial analysts are able to reduce information asymmetry between managers and investors (Frankel and Li, 2004; Healy and Palepu, 2001). Financial analysts are able to produce information about future earnings of the firm by relying on the firm’s accounting information (Barker and Imam, 2008; Salerno, 2014), which helps to uncover managements’ private information. Financial analysts of MNCs have to take into account financial information of all the subsidiaries and the parent to make forecasts about future earnings of the MNC (Herrmann, Inoue and Thomas, 2007). They have to gather information about both domestic earnings as foreign earnings of the MNC (Herrmann et al., 2007).
As prior studies (Dyreng, Hanlon and Maydew, 2012; Durnev et al., 2017) found that managers of MNCs prefer to engage in EM through their subsidiaries, it is likely that FRQ of subsidiaries affects the reporting quality of the MNC as a whole, and consequently the forecast accuracy of analysts. Therefore, I expect a positive relation between subsidiary FRQ and forecast accuracy. Additionally, prior researchers found that managers engage more in EM in countries with weaker IQ (Beuselinck, Cascino, Deloof and Van Straelen, 2018; Durnev et al., 2017). Hence, I expect that IQ of a country affects the relationship between subsidiaries FRQ and forecasts accuracy of analysts about the parent MNC negatively. This study investigates the effect of subsidiaries’ FRQ on forecasts of analysts about the parent company. Additionally, the effect of IQ of a subsidiaries’ country on the relationship between subsidiaries’ FRQ and forecasts of analysts is examined.
Specifically, this research investigates a sample of 312 U.S. MNCs which have in total 1,797 private European subsidiaries. Subsidiary FRQ is measured with an aggregate discretionary accruals measure calculated as the average discretionary accruals of all the subsidiaries controlled in a given year by a given parent, while the forecast accuracy is measured by the relative forecast error of the analysts following the parent company. The IQ is measured with an aggregate rule of law score taken for each country out of the WGI index.
This study used a sample of 12,344 forecasts of analysts and found that subsidiaries FRQ is positively related to the forecast accuracy of analysts. The IQ of a country, which is measured by the rule of law score, shows no effect on the relation between subsidiaries FRQ and forecast accuracy.
7
This research contributes to the literature by focusing on the relationship between FRQ and forecast accuracy. Previous studies found that FRQ is associated with the forecast accuracy of analysts (Mensah, Song and Ho, 2014; Salerno, 2014). These studies were focused on stand-alone companies in general and not specifically on MNCs. However, this is the first study that investigates this for MNCs and extends prior research by finding evidence that higher FRQ at subsidiary level increases the forecast accuracy of analysts following the parent MNC.
Besides, this study contributes to the literature which investigates the effect of IQ on FRQ. By considering IQ in countries of subsidiaries, prior literature (Beuselinck et al., 2018; Durnev et al., 2017) found evidence that FRQ is higher when IQ of a subsidiaries’ country is better. However, MNCs prefer to engage in EM in countries with weaker IQ (Beuselinck et al., 2018; Dyreng et al., 2012). The IQ of a subsidiaries’ country may influence the EM practices of managers of MNCs. Therefore, this will be the first study that investigated the positive effect of IQ on FRQ of the subsidiaries and therewith the forecast accuracy of analysts following the parent MNC. However, this study did not find that the IQ of a subsidiaries’ country affects the FRQ of the subsidiary and thus the forecast accuracy about the parent MNC positively or negatively.
The remainder of the paper proceeds as follows: In section 2, I review the literature and describe the hypothesis development. In section 3, I describe the sample selection, variable description and the empirical model. In section 4, I present the findings of the study. Section 5 contains the discussion and conclusion.
8
Literature
Literature review
Jonas and Blanchet (2000) describe FRQ as ‘full and transparent financial information that is not designed to obfuscate or mislead users. The IASB, is an international organization for accounting standards, which sets standards for firms to disclose full and transparent information that does not deceive users of the financial statements. The IASB suggests that the objective of financial reporting is ‘to provide financial information about the reporting entity that is useful to present and helps potential equity investors, lenders and other creditors in making decisions in their capacity as capital providers (IASB, 2008).
Previous studies (Jung, Lee and Weber, 2014; Rajgopal and Venkatachalam, 2011; Chen, Hope, Li and Wang, 2011) found that better FRQ could reduce the information asymmetry between managers and investors. They suggest that FRQ is negatively related with information asymmetry between the shareholders and the managers of the firm.
Additionally, prior research documents that companies benefit from reporting good quality accounting numbers. A group of researchers found that higher FRQ affects the economic performance of a firm positively. Firms who try to reduce information asymmetries between market participants by producing good information for share- and stakeholders, enjoy a more positive assessment of firm performance compared to firms that produce lower FRQ information (Martínez-Ferrero, 2014; Ahmed and Duellmand, 2011; Bushman and Smith, 2001).
Moreover, there is a stream of literature that investigated the relation between FRQ and cost of debt. Francis, LaFond, Olsson and Schipper (2005) found that firms with poorer earnings quality have to pay higher interest expenses, hence get less attractive debt contracting terms compared to firms with higher FRQ information. Poor quality reporting brings more information risk for investors, wherefore they choose to protect themselves by more stringent debt contracting terms (Leuz and Verrecchia, 2000). Thus, according to prior research (Francis et al., 2005; Leuz and Verrecchia, 2000) FRQ is negatively related to cost of debt.
Eventually, there is a group of studies which found evidence that FRQ is related to the forecasts of analysts following the firm. Burgstahler and Eames (2003) found that analysts anticipate on the FRQ of the firm in their forecasts made about the future performance of the firm. In addition, prior research found (Mensah et al., 2014; Salerno, 2014) that higher FRQ increases the accuracy of future financial
9
performance estimates of firms done by analysts. So, they suggest that FRQ is positively related with forecast accuracy of analysts about the firm.
Previous literature often uses earnings management to measure FRQ. Earnings management (EM) occurs when managers influence and alter financial reporting to either mislead stakeholders about the underlying economic performance or to influence contractual outcomes that rely on reported accounting numbers (Healy and Wahlen, 1999). Healy and Wahlen (1999) distinguish three incentive categories that motivate managers to engage in EM, namely (1) capital market expectations and valuation, (2) incentive contracts and (3) antitrust or other government regulation.
Financial accounting information is used by analysts and investors to value stock prices and make forecasts about the future performance of the firm. Hence, managers are incentivized to engage in EM to influence the stock prices and forecasts (Healy and Wahlen, 1999). The second category of incentives is based on lending and compensation contracts. Lending contracts frequently have debt covenants included, that do not allow the existence of reported losses. Reported losses rather want to be avoided due to high penalties for violation (Dichev and Skinner, 2002; Saleh and Ahmed, 2005). Compensation contracts frequently include earnings benchmarks in combination with earnings-based awards. Managers are more willing to engage in EM, because of the compensations they could gain by achieving benchmarks (Healy, 1985; Holthausen, Larcker and Sloan, 1995). Finally, Healy and Wahlen (1999) emphasize the regulatory incentives managers have. Managers could engage in EM to suffice regulatory requirements. Previous studies found that managers have incentives to manage earnings so that they comply with earnings thresholds of earnings-based listing regulation (Garcíchea‐M eca and Sanchez‐ Ballesta, 2009; Cheng, Aerts, Jorissen, 2010).
Graham (2005) conducted a survey among CFOs, which revealed that analysts’ forecasts is amongst manager’s most relevant benchmarks. Managers who do not meet analysts’ forecasts frequently experience high penalties by market (Lopez and Rees, 2002). The compensation of managers is based on benchmarks as the analysts’ forecasts and share prices, which increases the pressure for managers to meet or beat them. Additionally, firms who not meet analysts’ forecasts, experience lower stock prices compared to firms who succeed to meet or beat analysts’ forecasts (Bartov, Givoly and Hahn, 2002). So, the analysts’ forecasts are an important benchmark for managers. They help to get an overview of a firm’s performance (Yu, 2010). The financial statement information is used by analysts to make forecasts (Yu, 2010).
10
The financial statement information of firms is not always as useful for analysts’ forecasts. McVay (2006) found that managers start to engage in activities as EM, classification shifting, real activities manipulation and/or forecast guidance when there is an opportunity to meet or beat analyst forecasts (McVay, 2006). The financial statement information could be influenced by these activities. Therefore, it is important to address the different types of roles analysts have in the market. Lang and Lundholm (1996) distinguish two types of roles analysts have in the capital market. Financial analysts could have the role of an information producer or an information intermediary. An information producer is able to produce relevant information for the market by themselves, while an information intermediary only passes the correct information on to the market (Lang and Lundholm, 1996). Analysts sell their reports about future financial performance of the firm to the market. The information producer analysts’ report includes information with own produced relevant information about functioning of the firm and competes with the firm-provided disclosure. On contrary, the information intermediary analysts’ report only includes information originally coming from the financial statement information of the firm and is more a complement of the firm-provided disclosure. Analysts as information producers will face a decrease in sales when the financial statement information of the firm increases.
Lang and Lundholm (1996) argue that an increase in FRQ, thus lower levels of EM results in less valuable forecasts to sell. The reports of the analysts become less useful, because there is less opportunity to show their value with their reports by giving information about EM practices of managers. When the EM practices increase, the financial statement information decreases, which consequently increases the value of the reports from the information producers. However, analysts who fulfill the role of information intermediaries, face a decrease in sales of their reports when EM practices increase for the firm. This is because they are not able to capture EM practices of managers as precise as the information producers are able to.
11
Hypothesis development
Lang and Lundholm (1996) investigated the effect of information disclosure on analyst following, forecast volatility, forecast dispersion and forecast accuracy and found evidence that more disclosure of information could improve the accuracy of analysts’ forecasts. Moreover, there is evidence that higher disclosure quality leads to more accurate forecasts by analysts (Hope, 2003; Vanstraelen, Zarzeski and Robb, 2003; Eng and Teo, 1999). This is due to the fact that firms with high quality levels of disclosure will engage less in earnings management than firms with lower levels of disclosure (Lobo and Zhou, 2001). Additionally, prior research investigated if there is a direct relationship between EM and analysts and found that managers who engage in EM and therewith manipulate the annual reports, negatively impact the accuracy of analysts’ current year earnings forecasts (Wilson and Wu, 2011; Salerno, 2014; Embong and Hosseini, 2018).
Prior research found that corporate international diversification leads to more information asymmetry between managers and investors, which makes it more difficult for analysts to make accurate forecasts about future financial performance of the firm (Chin, Chen and Hsieh, 2009; Ashbaugh and Pincus, 2001). MNCs have higher corporate international diversification, and therefore more information asymmetry between managers and investors, which gives managers of MNCs more opportunities to engage in EM than those of domestic firms (Chin et al., 2009). Additionally, Fan (2012) found that MNCs manage earnings to avoid losses through foreign earnings and not through domestic earnings. The higher complexity of foreign earnings and the fewer information available for investors creates more flexibility for managers with offshore firms (Durnev et al., 2009; Fan, 2012).
Previous research found that managers of MNCs have a preference for offshore firms to manage earnings. Managers of MNCs engage in EM at subsidiaries countries with lower investor protection, less strict rule of law and tax-haven status (Dyreng, et al., 2012; Durnev et al., 2017; Beuselinck et al., 2018). Besides, prior research found that MNCs engage in EM at the subsidiaries where they have sufficient influence (Beuselinck et al., 2018). The subsidiaries have to be consolidated in the financial statements of the MNCs and therefore may be used by MNCs to meet their consolidated reporting objectives (Beuselinck et al., 2018). The analysts’ forecasts based on the consolidated statements of the MNC could be influenced by the EM that happens at the subsidiary level of the MNC. The higher FRQ, thus less EM, will make it simpler for analysts to predict future earnings. This is supported by previous studies, who found that companies with higher FRQ experience higher forecast accuracy by analysts, and vice versa (Wilson and Wu, 2011; Salerno, 2014; Embong and Hosseini, 2018). Since higher EM
12
at the subsidiary level decreases the FRQ of the MNC as a whole, I expect that analysts’ forecasts made about the consolidated statements of the MNC will be more accurate if the FRQ at subsidiary level is higher. This leads to the following hypothesis:
H1 = Subsidiary FRQ is positively associated with the forecast accuracy of analysts following the MNC.
Greif (2006) defines institutions as ‘a set of social factors, rules, beliefs, values and organisations that jointly motivate regularity in individual and social behaviour’. Good institutions help to shape human behaviours by providing a system of shared beliefs (Alonso and Garcimartin, 2013). The World Bank (2002) argues that the IQ of a country depends upon the quality of its rules, the enforcement procedures it has in place to encourage society to comply with these rules, and the performance and guidance available for firms. Previous research (Dollar, Hallward-Driemier and Mengistae, 2005) found that the IQ of a country affects many participants of society in a country, among which the firms in a country. They found that the IQ of a country is positively related to firm performance. Additionally, prior studies investigated the relationship between country’s IQ and the FRQ of a firm. A group of researchers (Ball, Kothari and Robin, 2000; Giner and Rees, 2001; Leuz, Nanda and Wysocki, 2003; and Bushman and Piotroski, 2006) found a positive relationship between regulatory environment, timeliness and conservative accounting. They found that firms which are incorporated in countries with stronger regulatory government systems recognize their earnings timelier and are more conservative by incorporating bad news in earnings earlier than in countries with weaker government systems. Besides, previous studies (Burgstahler, Hail and Leuz, 2006; Van Tendeloo and Vanstraelen, 2008) found a positive relationship between a country’s rule of law and FRQ. Higher investor protection mechanisms are associated with less prevalent EM practices from the managers of firms.
Then, there is stream of literature which investigated the role of IQ for FRQ of MNCs. Previous researchers found that IQ relates to the FRQ for MNCs (Durnev et al., 2017; Dyreng, 2012; Beuselinck et al., 2018). Durnev et al. (2017) finds that firms with offshore investments have lower FRQ than firms without offshore investment centers. Dyreng et al. (2012) found that geographical location affects earnings management, due to the rule of law and the tax-haven status that different countries have (Dyreng, Hanlon and Maydew, 2012). The lower rule of law and the tax-haven status of a country relate to more earnings management practices at the subsidiary level compared to subsidiaries that are incorporated in higher rule of law countries and countries that do not have a tax-haven status. Beuselinck et al. (2018) found that MNCs manage their consolidated earnings throughout their subsidiaries. MNCs
13
who have established their headquarters in a country with a strong rule of law, for example the U.S, manage their earnings more via subsidiaries in countries with a weaker rule of law.
These studies show evidence that managers of MNCs manage their earnings through their subsidiaries. The weaker regulatory systems and tax systems create these opportunities for managers. This study is focused on MNCs that are headquartered in the U.S. The U.S. is known to have a strict regulatory environment, including stringent enforcement rules and possible class action lawsuits (Huijgen and Lubberink, 2005). Due to a strict regulatory environment as in the U.S., MNCs might prefer to manage earnings abroad, especially in countries with weaker regulatory environments (Durnev et al., 2017). Lower institutional quality in countries is related to more EM practices and thus lower FRQ (Dyreng, et al., 2012; Durnev et al., 2017; Beuselinck et al., 2018). If this is the case,the IQ of the countries where subsidiaries are located might strengthen the relationship between subsidiary’s FRQ and forecast accuracy of analysts following the MNC. Higher IQ might influence the subsidiaries FRQ positively and thereby strengthen the relationship between subsidiaries FRQ and forecast accuracy.
The following hypothesis is constructed:
H2 = The relationship between the subsidiary’s FRQ and forecast accuracy of analysts of the MNC is strengthened when IQ of country of the subsidiary is higher.
14
Methodology
Sample selection
The sample selection started by identifying non-financial public MNCs that are incorporated in the U.S. during the period 2015-2017. For each MNC I hand-collected the list of material subsidiaries disclosed in exhibit 21 of form 10-K from the Edgar Portal. Given the availability of the data, I selected only the non-financial private subsidiaries located in 30 European countries, made up of 28 EU countries, Norway, and Switzerland, and obtained their financial information from the Orbis database. Financial information of the US MNCs is obtained from the Compustat database, and all information regarding the analyst’s forecasts comes from the I/B/E/S database. The final sample consists of 12,344 forecasts over the period 2015 until 2017, whereby information about 1,797 subsidiaries is included that originate from the 312 MNCs.
Forecast accuracy
Consistent with previous studies (Barniv et al., 2005; Salerno, 2014), I use the relative forecast error (RFE) of analysts to measure the forecast accuracy. The RFE formula is as follows:
𝑅𝑅𝑅𝑅𝑅𝑅 = �𝐴𝐴𝐴𝐴𝐸𝐸𝑖𝑖𝑖𝑖𝑖𝑖
𝑀𝑀𝐴𝐴𝐴𝐴𝐸𝐸𝑖𝑖𝑖𝑖� − 1 (1)
AFE is the absolute forecast error, calculated as the absolute value of the difference between the Forecasted EPSi,t,j and the Actual EPSi,t,j. Forecast EPS is analyst i’s most recent forecast of the
earnings per share in the twelve months before the earnings announcement for MNC j at time t. The
Actual EPS i,t,j is the reported earnings per share, where i is the individual analyst, t is the fiscal year
of the reported EPS and j is the MNC in issue. MAFE is the Mean Absolute Forecast error of the analyst, which is calculated by computing the mean between the Forecasted EPSi,t and Actual EPSi,t ,where i
is all analysts’ consensus forecast and t the year wherein the forecasts have been made.
Eventually, RFE is calculated by dividing AFE by MAFE, whereby the ratio will be subtracted by 1 and multiplied by -1. This creates a ratio, wherein a score above 0 means more accurate than the average forecast accuracy and a negative score means a forecast accuracy that is below average.
15
Financial reporting quality
I use accrual based earnings management to measure subsidiary FRQ, which is widely measured with the magnitude of the discretionary accruals (Dyreng et al., 2012; Salerno, 2014; Beuselinck et al., 2018). The most common discretionary accrual models are the Jones model (1991) and the modified Jones model by Dechow, Sloan and Sweeney (1995). The modified Jones model is preferred over the Jones model because the change in revenues is adjusted for the change in receivables in the event period. The problem with the Jones model is that the revenues are still included in the estimate of the non-discretionary accruals (Dechow et al., 1995). Therefore, the revenue could be manipulated by managers which influences the level of non-discretionary accruals. By adjusting the revenues for the receivables in the event period the error will be removed and the earnings discretion can be measured. Therefore, this study uses the modified Jones model.
First, I will use the Jones model (1991) to calculate the total accruals for the modified Jones model.To calculate the total accruals of the formula, the equation mentioned below will be used:
𝑇𝑇𝐴𝐴𝑡𝑡 = ∆𝐶𝐶𝐴𝐴𝑡𝑡− ∆𝐶𝐶𝐶𝐶𝐶𝐶ℎ𝑡𝑡− ∆𝐶𝐶𝐿𝐿𝑡𝑡+ ∆𝑆𝑆𝑇𝑇𝑆𝑆𝑅𝑅𝑆𝑆𝑇𝑇𝑡𝑡− 𝑆𝑆𝑅𝑅𝑃𝑃𝑡𝑡 (2)
Where, ∆𝐶𝐶𝐴𝐴𝑡𝑡 is the current assets in year t minus current assets in year t-1, ∆𝐶𝐶𝐶𝐶𝐶𝐶ℎ𝑡𝑡 is the cash and cash
equivalents in year t minus current assets in year t-1, ∆𝐶𝐶𝐿𝐿𝑡𝑡 is the current liabilities in year t minus the
current liabilities in year t-1, 𝑆𝑆𝑇𝑇𝑆𝑆𝑅𝑅𝑆𝑆𝑇𝑇𝑡𝑡 is the short term debt in year t minus short term debt in year t-1
and 𝑆𝑆𝑅𝑅𝑃𝑃𝑡𝑡 is the depreciation and amortization expenses in year t.
Consequently, I will use the modified jones model to identify the discretionary accruals for each industry-year-subsidiary combination. 𝑇𝑇𝐴𝐴 𝑖𝑖 𝐴𝐴𝑖𝑖−1= 𝛼𝛼1∗ � 1 𝐴𝐴𝑖𝑖−1� + 𝛼𝛼2∗ � ∆𝑅𝑅𝐸𝐸𝑅𝑅𝑖𝑖 − ∆𝑅𝑅𝐸𝐸𝑅𝑅𝑖𝑖 𝐴𝐴𝑖𝑖−1 � + 𝛼𝛼3∗ � 𝑃𝑃𝑃𝑃𝐸𝐸𝑖𝑖 𝐴𝐴𝑖𝑖−1� + 𝜀𝜀𝑡𝑡 (3)
Where 𝑇𝑇𝐴𝐴𝑡𝑡 , is the total accruals in year t, ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡 is the change in revenues in year t, ∆𝑅𝑅𝑅𝑅𝐶𝐶𝑡𝑡 is the
change in receivables in year t and 𝑃𝑃𝑃𝑃𝑅𝑅𝑡𝑡 is the property, plant and equipment in year t. The variables
in the model are divided by the lagged value of the total assets (𝐴𝐴𝑡𝑡−1) of the subsidiary and 𝜀𝜀𝑡𝑡 are the
16
The calculation of the non-discretionary accruals is necessary to calculate the discretionary accruals of the subsidiaries. Therefore, I use the following formula:
𝑁𝑁𝑆𝑆𝐴𝐴𝑆𝑆𝑆𝑆𝑆𝑆 = 𝛼𝛼1�𝐴𝐴𝑖𝑖−11 � + 𝛼𝛼2�∆𝑅𝑅𝐸𝐸𝑅𝑅𝐴𝐴𝑖𝑖 − ∆𝑅𝑅𝐸𝐸𝑅𝑅𝑖𝑖−1 𝑖𝑖� + 𝛼𝛼3�𝑃𝑃𝑃𝑃𝐸𝐸𝐴𝐴𝑖𝑖−1𝑖𝑖� (4)
Finally, the discretionary accruals can be calculated to measure the FRQ of the subsidiary. The following formula is used:
|𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶|𝑆𝑆𝑆𝑆𝑆𝑆 = 𝐴𝐴𝑇𝑇𝐴𝐴𝑖𝑖−1𝑖𝑖 −(𝐴𝐴𝑁𝑁𝑁𝑁𝐴𝐴𝑖𝑖−1𝑖𝑖 (5)
The discretionary accruals of subsidiaries are calculated by using an aggregate absolute measure to calculate the discretionary accruals (|𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶|𝑆𝑆𝑆𝑆𝑆𝑆) of the subsidiaries of an MNC. The aggregate measure
is calculated by taking the sum of the absolute value of the discretionary accruals of all the MNCs subsidiaries divided by the number of subsidiaries for the specific MNC.
Institutional quality
The WGI (Worldwide Governance Indicators) database created by the World Bank has been used in previous studies (Dyreng et al., 2012; Beuselinck., 2018) to measure the IQ of a country. The World Bank constructed a database, whereby the IQ of a country is measured by six dimensions. The data in the database is obtained from a large number of enterprises, citizens, and expert survey respondents from countries over the whole world. Following prior studies who investigated the effect of IQ of subsidiaries countries (Dyreng et al., 2012; Beuselinck et al., 2018), this study uses the RULE OF
LAWSUB, defined as the annual rule of law index scores by country, as a proxy of the IQ of the countries
where subsidiaries are located. The rule of law variable (RULE OF LAWSUB) is an aggregate measure of
all the subsidiaries in the sample that belong to a specific MNC. The aggregate measure is used to measure the IQ of the subsidiaries that belong to the MNC.
17
Empirical model
I will use an ordinary least squares regression (OLS) to test both hypothesis. To test hypothesis one, I estimate the following model:
𝑅𝑅𝑅𝑅𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅 = 𝛼𝛼 + 𝛽𝛽1|𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶|𝑆𝑆𝑆𝑆𝑆𝑆+ 𝛽𝛽2𝑅𝑅𝑅𝑅𝐴𝐴𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽3𝑅𝑅𝑅𝑅𝑅𝑅𝑄𝑄𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽4𝐶𝐶𝑅𝑅𝐶𝐶𝑃𝑃𝑀𝑀𝑁𝑁𝑅𝑅+
𝛽𝛽5𝑆𝑆𝑆𝑆𝑆𝑆𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽6𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽7𝑅𝑅𝐴𝐴𝑅𝑅𝑁𝑁𝑆𝑆𝑁𝑁𝐸𝐸𝑆𝑆_𝑆𝑆𝑆𝑆𝑅𝑅𝑃𝑃𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽8𝐻𝐻𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑁𝑁𝑀𝑀𝑁𝑁𝑅𝑅+
𝛽𝛽9𝑁𝑁𝐴𝐴𝑅𝑅𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽10𝑆𝑆𝑁𝑁𝑆𝑆𝑆𝑆𝑆𝑆𝑇𝑇𝑅𝑅𝑌𝑌𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽11𝑌𝑌𝑅𝑅𝐴𝐴𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅+ 𝜀𝜀 (6)
Given my prediction that the relative forecast error (𝑅𝑅𝑅𝑅𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅) increases when the aggregate absolute
discretionary accruals increase, the coefficient of |DACC|SUB is expected to be positive.
To test hypothesis 2, I extend the baseline model with two regressors. These are the rule of law (𝑅𝑅𝑆𝑆𝐿𝐿𝑅𝑅 𝑅𝑅𝑅𝑅 𝐿𝐿𝐴𝐴𝑊𝑊𝑆𝑆𝑆𝑆𝑆𝑆) and the interaction term between rule of law and discretionary
accruals( 𝑅𝑅𝑆𝑆𝐿𝐿𝑅𝑅 𝑅𝑅𝑅𝑅 𝐿𝐿𝐴𝐴𝑊𝑊 𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶𝑆𝑆𝑆𝑆𝑆𝑆∗ |𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶|𝑆𝑆𝑆𝑆𝑆𝑆). 𝑅𝑅𝑆𝑆𝐿𝐿𝑅𝑅 𝑅𝑅𝑅𝑅 𝐿𝐿𝐴𝐴𝑊𝑊𝑆𝑆𝑆𝑆𝑆𝑆 is an aggregate measure for the
IQ of the subsidiary. The interaction term (𝑅𝑅𝑆𝑆𝐿𝐿𝑅𝑅 𝑅𝑅𝑅𝑅 𝐿𝐿𝐴𝐴𝑊𝑊 |𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶|𝑆𝑆𝑆𝑆𝑆𝑆∗ |𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶|𝑆𝑆𝑆𝑆𝑆𝑆) is used to
measure the moderating effect on the relationship between FRQ and forecast accuracy of analysts. The equation to test the second hypothesis is as follows:
𝑅𝑅𝑅𝑅𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅 = 𝛼𝛼 + 𝛽𝛽1|𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶|𝑆𝑆𝑆𝑆𝑆𝑆+ 𝛽𝛽2𝑅𝑅𝑆𝑆𝐿𝐿𝑅𝑅 𝑅𝑅𝑅𝑅 𝐿𝐿𝐴𝐴𝑊𝑊𝑆𝑆𝑆𝑆𝑆𝑆+ 𝛽𝛽3𝑅𝑅𝑆𝑆𝐿𝐿𝑅𝑅 𝑅𝑅𝑅𝑅 𝐿𝐿𝐴𝐴𝑊𝑊 ∗
|𝑆𝑆𝐴𝐴𝐶𝐶𝐶𝐶|𝑆𝑆𝑆𝑆𝑆𝑆+ 𝛽𝛽4𝑅𝑅𝑅𝑅𝐴𝐴𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽5𝑅𝑅𝑅𝑅𝑅𝑅𝑄𝑄𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽6𝐶𝐶𝑅𝑅𝐶𝐶𝑃𝑃𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽7𝑆𝑆𝑆𝑆𝑆𝑆𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽8𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆𝑀𝑀𝑁𝑁𝑅𝑅+
𝛽𝛽9𝑅𝑅𝐴𝐴𝑅𝑅𝑁𝑁𝑆𝑆𝑁𝑁𝐸𝐸𝑆𝑆_𝑆𝑆𝑆𝑆𝑅𝑅𝑃𝑃𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽10𝐻𝐻𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑁𝑁𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽11𝑁𝑁𝐴𝐴𝑅𝑅𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅+
𝛽𝛽12𝑆𝑆𝑁𝑁𝑆𝑆𝑆𝑆𝑆𝑆𝑇𝑇𝑅𝑅𝑌𝑌𝑀𝑀𝑁𝑁𝑅𝑅+ 𝛽𝛽13𝑌𝑌𝑅𝑅𝐴𝐴𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅+ 𝜀𝜀 (7)
Given my prediction that the relationship between the subsidiaries FRQ and forecast accuracy of analysts about the parent MNC is strengthened by higher IQ, I expect that the rule of law (𝑅𝑅𝑆𝑆𝐿𝐿𝑅𝑅 𝑅𝑅𝑅𝑅 𝐿𝐿𝐴𝐴𝑊𝑊𝑆𝑆𝑆𝑆𝑆𝑆) strengthens the relationship between aggregate discretionary accruals (|DACC|SUB)
and the relative forecast error (𝑅𝑅𝑅𝑅𝑅𝑅𝑀𝑀𝑁𝑁𝑅𝑅) negatively.
Control variables
Previous literature documents the existence of a relationship between several variables and forecast accuracy. I checked these variables in the models to test the hypothesis. In accordance with Beuselinck et al. (2018) I monitored the performance (ROAMNC) of the MNC, measured by the value of the return on
18
firm. Following Beuselinck et al. (2018) I expect that firms that are performing better will have a higher return on assets and will consequently have a higher forecast accuracy. Besides, I control in the models for the frequency of analysts’ forecasts (FREQMNC). This is a proxy for the amount of effort an analyst
dedicates to a firm (Clement and Tse, 2003), calculated by the number of forecasts done by analyst i for MNC j at time t. Jacob and Neale (1999) discovered evidence that the forecast accuracy increases by a rise in the number of forecasts done by an analyst Investing more effort in the forecasts that analysts make about a specific firm could result in more accurate forecasts of the firm (Clement and Tse, 2003). Following prior studies (Jacob and Neale, 1999; Clement and Tse, 2003 Salerno, 2014; Beuselinck et al., 2018), I expect that the frequency of analysts is positively associated with the forecast accuracy of the analysts.
Lang and Lundholm (1996) found evidence that the firm size (SIZEMNC) is associated with the forecast
accuracy of analysts. They found that larger firms have a higher forecast accuracy compared to firms with a smaller size. For this reason, I include the firm size (SIZEMNC) as a control variable in the models.
The firm size (SIZEMNC) is calculated as the natural log of the total assets for MNC j at time t, and
positively affects the forecast accuracy of analysts. Hwang, Jan and Basu (1996) discovered that the firms that report a loss in a certain year have on average less accurate forecasts than firms with higher profits. This is the case because loss firms are mainly followed by sell-side analysts, who on average over predict earnings (Hwang et al., 1996). Therefore, I include a dummy variable for loss of the MNC (LOSSMNC) that is 1 if the MNC has negative earnings, 0 otherwise.
Besides, I control for the earnings surprise (EARNINGS_SURPRISEMNC), calculated by the difference
between current year’s EPS and last year’s EPS, divided by the price at the beginning of the fiscal year. Lang and Lundholm (1996) found that earnings surprise affects the forecast accuracy because larger changes in earnings correspond with less accurate forecasts. Hence, I expect that higher earnings volatility leads to less accurate forecasts. Following prior studies relating to forecast accuracy (Barniv and Thomas, 2005; Ben, Choi and Kang, 2008; Luo, 2009 and Salerno, 2014) the forecast horizon (HORIZMNC) is used in the model as control variable. The forecast horizon (HORIZMNC) is calculated by
the total number of days a forecast is made before the earnings announcement date, divided by 365. It is expected that a forecast closer to the announcement date is more accurate than a forecast more beyond the announcement date (Ben et al., 2008) because prior literature found that analysts are more accurate over shorter horizons (Barniv and Thomas, 2005; Ben et al., 2008; Luo, 2009 and Salerno, 2014). Following Hope (2003) and Salerno (2014), I also monitored for the number of analysts following a firm (NAFFMNC), calculated as the number of analysts following a specific MNC. They found a positive
19
relation between the quantity of analysts following and the forecast accuracy.Firms with more analysts following the firm have a higher forecast accuracy compared to firms with less analysts following the firm (Hope, 2003; Salerno, 2014). Additionally, I checked for the analyst complexity (COMPMNC),
measured by the total number of MNCs j that the analyst i follows at time t. In accordance with prior studies (Barniv and Thomas, 2005; Salerno, 2014) that found that higher complexity for analysts is associated with lower forecast accuracy, I expect a negative relationship between the analyst complexity and forecast accuracy.
Lastly, I included year and industry effects. I used year dummies to control for the time series effects and I used industry dummies based on 1-digit SIC codes of the industry that the MNC operates in.
20
Results
Descriptive statistics
Table 1 provides the descriptive statistics for all the variables. The dependent variable, relative forecast error (RFEMNC), has a mean of 0.030, a minimum of -0.170 and a maximum of 0.296. The forecast of
the analyst with an RFE of -0.170 means that the forecast was far from the actual EPS, while an RFE score of 0.296 means that the forecast was closer to the actual EPS. The main independent variable, which is the absolute value of the aggregate discretionary accruals of subsidiaries of MNCs (|DACC|SUB),
has a mean of 0.020 and a standard deviation of 0.024. The rule of law score (RULE_OF_LAWSUB) has
a mean of 1.089, which means that average rule of law is relatively good in the sample. This is not surprising because the subsidiaries of the MNCs analyzed are all based in European countries.
Further, the frequency variable (FREQMNC) has an average of 4, which means that the average forecasts
made by an analyst for a specific MNC is 4 in a fiscal year. Besides, the minimum number of forecasts done for a specific MNC is 1 and the maximum is 11. The variable complexity (COMPMNC) shows that
an analyst follows, on average, 7 MNCs in a year and maximum 19 MNCs. The ROA (ROAMNC) has a
mean of 0.074, which means that the average firm has a positive return in value on their assets. The average size of the MNC (TOTAL_ASSETSMNC) in the sample is $18,337,020. Loss (LOSSMNC) has a mean
of 0.145, this means that most of the MNCs report a negative result in the specific year. The surprise in earnings (EARNINGS_SURPRISEMNC) is on average negative (-0,007). A negative average earnings
surprise means that the earnings are more negative than the analysts expected on forehand. The forecast horizon (HORIZONMNC) is on average 0.234 indicating that the average forecast is made 85 days before
the actual earnings are announced. The average number of analysts following an MNC (NAFFMNC) is
21 This table shows the descriptive statistics of all the variables used in the model, whereby all continuous variables are winsorized at a 10% level. The sample size, mean, median, standard deviation, minimum and maximum are included. The measure for the FRQ (|DACC|SUB), is the aggregate measure for the absolute value of the discretionary accruals of the
subsidiary of the MNC. The aggregate measure is calculated by taking the sum of the absolute value of the discretionary accruals of all the MNCs subsidiaries divided by the number of subsidiaries for the specific MNC. The rule of law (RULE_OF_LAWSUB) is an aggregate measure of the rule of law, which is one of the WGI indicators and provides the score
of a country’s rule of law quality (Kaufman et al., 2009). The aggregate measure is calculated by taking the sum of the absolute value of the rule of law scores of all the MNCs subsidiaries divided by the number of subsidiaries for the specific MNC. The relative forecast error (RFEMNC) is the measure for the forecast accuracy of the analyst about the parent of the
MNC. The relative forecast error is calculated by dividing AFE by MAFE, whereby the ratio is subtracted by 1 and multiplied by -1. For calculation of the AFE and the MAFE, see the methodology section. ROAMNC is the value of return on assets,
calculated as the net income before extraordinary items divided by the total assets of a firm. FREQMNC is the number of
forecasts done by an analyst for a specific MNC each year. The COMPMNC is the number of MNCs forecasted by an analyst.
The TOTAL_ASSETSMNC are the total assets of the MNC parent in millions respectively. LOSSMNC is a dummy equal to 1 if
the MNC reports a loss in a specific year, 0 otherwise. EARNINGS_SURPRISEMNC is earnings of an MNC that are above
analyst’s expectations. HORIZONMNC is the forecast time before a MNCs earnings announcement. The NAFFMNC is the
number of analysts following a specific MNC each year.
Table 1 : Descriptive Statistics
Variable n Mean Median Std Dev Minimum Maximum
Subsidiary-level variables: |DACC|SUB 12,344 0.020 0.011 0.024 0.000 0.129 RULE_OF_LAWSUB 12,344 1.089 1.089 0.330 0.275 2.020 Parent-level variables: RFEMNC 12,344 0.030 0.005 0.129 -0.170 0.296 ROAMNC 12,344 0.074 0.059 0.060 0.002 0.351 FREQMNC 12,344 4.397 4 2.222 1 11 COMPMNC 12,344 6.745 6 3.640 1 19 TOTAL_ASSETSMNC 12,344 18,337,020 5,695,800 33,180,753 53,916 221,690,000 (U.S. $000) LOSSMNC 12,344 0.145 1 0.352 0 1 EARNINGS_SURPRISEMNC 12,344 -0.007 0.001 0.071 -0.376 0.221 HORIZONMNC 12,344 0.234 0.170 0.194 0.022 0.910 NAFFMNC 12,344 19.80 20 9.593 3 45
22
Correlations
Table 2 shows the correlations between all the variables. Spearman correlations are reported below the diagonal and Pearson correlations are shown above the diagonal. The correlations are shown with significance at a 1%, 5% and 10% level. Results indicate that there is a negative correlation between the aggregate measure of subsidiary discretionary accruals (|DACC|SUB) and the relative forecast error of
analysts following the MNC (RFEMNC). The Pearson test results in a correlation coefficient of -0.029
with a 1% significance level, while the Spearman test results in a correlation coefficient of -0.019 with a 5% significance level. These coefficients indicate that when subsidiary discretionary accruals are lower, the analysts forecast accuracy is higher. Besides, the results show that the average rule of law of a country (RULE OF LAWSUB) has a correlation with the relative forecast error (RFEMNC), respectively
-0.009 under Pearson and -0.018 under Spearman. The former one is not significant, the latter is significant at a 5% level, suggesting that analyst forecasts are more accurate when the average rule of law of the MNC´s subsidiaries is lower. The rule of law of a country (RULE OF LAWMNC) is not
correlated with the discretionary accruals (|DACC|SUB), because the results show no significant
relationships.
The return on assets (ROAMNC) and the number of forecasts done by an analyst (FREQMNC) show both a
significant positive relation with the forecast accuracy at a 1% significance level. These findings show for both tests and indicate that the forecasts of analysts are more accurate. The results for return on assets show under Pearson respectively 0.073 and under Spearman 0.048. The findings for number of forecasts done by an analyst show respectively 0.100 under Pearson and 0.120 under Spearman.
Under the Pearson test, analyst complexity (COMPMNC) and size of the MNC (SIZEMNC) show weaker
correlations with the relative forecast error (RFEMNC). The analyst complexity is in neither test
significantly related with forecast accuracy, which indicates that there is no relation between analyst complexity and forecast accuracy. The size of the MNC is negatively significant under the Pearson test, -0.028 with a 1% significance level. On contrary, the Spearman test shows no relationship between size of the MNC and the forecast accuracy.
The loss of a MNC (LOSSMNC) and the earnings surprise (EARNINGS_SURPRISEMNC) show mixed
results for the correlation with relative forecast error (RFEMNC). The loss of an MNC in relation to the
relative forecast error shows under the Spearman test a correlation of -0.019 at a 1% significance level. This indicates that there is a correlation between loss of an MNC and the forecast accuracy. However, the Pearson test shows no relationship between the loss of an MNC and the forecast accuracy. The
23
earnings surprise of an MNC shows a significant relation of 0.028 under the Pearson test at a 1% significance level, while the Spearman test shows no relationship between the earnings surprise of an MNC and the forecast accuracy of analysts.
The forecast horizon is significantly negative related to the forecast accuracy, by respectively -0.014 for the Pearson test and -0.117 for the Spearman test, both significant at a 1% level. This indicates that there is a negative correlation between forecast horizon and the forecast accuracy. The number of analysts following an MNC (NAFFMNC) is not related to the forecast accuracy under the Pearson test, while the
Spearman test shows a significant correlation of -0.024 at a 1% level. The results for the Pearson test indicate that the number of analysts following an MNC is not correlated with the forecast accuracy of the analysts. The Spearman test shows evidence that there is a negative significant relation between the variables.
25 This table represents the correlations between the independent variables and dependent variable by a Pearson- and a Spearman test. Pearson correlations are shown above the diagonal, Spearman correlations are shown under the diagonal. All continuous variables are winsorized at a 10% level. The *,**,***, indicates the level of statistical significance at 10%, 5% and 1% levels for the relationships. The relative forecast error (RFEMNC) is the measure for the forecast accuracy of the analyst about the parent of the
MNC. The relative forecast error is calculated by dividing AFE by MAFE, whereby the ratio is subtracted by 1 and multiplied by -1. For calculation of the AFE and the MAFE, see the methodology section. The measure for the FRQ (|DACC|SUB), this is the aggregate measure for the absolute value of the discretionary accruals of the subsidiary of the
MNC. The aggregate measure is calculated by taking the sum of the absolute value of the discretionary accruals of all the MNCs subsidiaries divided by the number of subsidiaries for the specific MNC. The rule of law (RULE_OF_LAWSUB) is an aggregate measure of the rule of law, which is one of the WGI indicators and provides the score
of a country’s rule of law quality (Kaufman et al., 2009). The aggregate measure is calculated by taking the sum of the absolute value of the rule of law scores of all the MNCs subsidiaries divided by the number of subsidiaries for the specific MNC. ROAMNC is the absolute value of return on assets is calculated as the net income before extraordinary
items divided by the total assets of a firm. FREQMNC is the number of forecasts done by an analyst for a specific MNC each year. The COMPMNC is the number of MNCs
forecasted by an analyst. SIZEMNC is the natural logarithm of the book value of the MNC’s total assets. LOSSMNC is a MNC which reports a loss in a specific year.
EARNINGS_SURPRISEMNC is earnings of a MNC that are above analyst’s expectations. HORIZONMNC is the forecast time before a MNCs earnings announcement. The
NAFFMNC is the number of analysts following a specific MNC each year.
Table 2: Pearson/Spearman Correlations
Variables A B C D E F G H I J K A: RFEMNC -0.029*** -0.009 0.073*** 0.100*** -0.003 -0.028*** 0.001 0.028*** -0.014*** -0.006 B: |DACC|SUB -0.019** 0.013 -0.020** -0.015 -0.014 -0.028*** -0.009 0.013 -0.002 -0.061*** C: RULE_OF_LAWSUB -0.018** 0.006 -0.131*** 0.005 -0.003 0.035*** -0.090*** 0.004 -0.017* 0.094*** D: ROAMNC 0.048*** -0.014 -0.134*** 0.026*** -0.044*** -0.050*** 0.008 -0.119*** 0.041*** 0.114*** E: FREQMNC 0.120*** -0.013 0.016 0.026*** 0.050*** 0.145*** -0.014 -0.027*** -0.470*** 0.159*** F: COMPMNC 0.011 -0.004 -0.010 -0.024*** 0.091*** -0.068*** -0.024*** -0.001 -0.107*** -0.053*** G: SIZEMNC 0.005 -0.045*** 0.016 0.048*** 0.128*** -0.088*** 0.143*** 0.027*** 0.008 0.654*** H: LOSSMNC -0.019** -0.025*** -0.104*** 0.164*** -0.010 -0.027*** 0.143*** 0.317*** 0.003 0.025*** I: EARNINGS_SURPRISEMNC -0.004 -0.010 -0.029*** 0.160*** -0.041*** -0.008 -0.021 0.250*** 0.010 -0.056*** J: HORIZONMNC -0.117*** 0.015 -0.053*** 0.088*** -0.357*** -0.036*** 0.098*** 0.080*** 0.023*** 0.025*** K: NAFFMNC -0.024*** -0.066*** 0.065*** 0.194*** 0.134*** 0.073*** 0.684*** 0.052*** -0.096*** 0.115***
26
Results regression analysis
I performed an OLS- regression to investigate the relationship between subsidiaries FRQ and the analyst’s forecast accuracy of the parent of the MNC. Therefore, I test the relationship between the aggregate discretionary accruals of the subsidiaries (|DACC|SUB) and the relative forecast error of the
analyst for the parent of the MNC (RFEMNC). The first model tests hypothesis 1, according to which I
expect a negative relationship between discretionary accruals and relative forecast error. The results show a significant coefficient of -0.217 at a 5% level for the relationship between those variables. This indicates that a lower subsidiary FRQ is associated with a decrease of 0.217 in forecast accuracy of the analyst that is evaluating the parent MNC. The results provide evidence to accept hypothesis 1.
The ROAMNC and FREQMNC , show a positive significant relationship with the RFEMNC, which suggests
that a higher return on assets and more forecasts by analysts is related to higher forecast accuracy, respectively 0.340 for return on assets and 0.004 for the forecast frequency. Both are significant at a 1% level. The results are in accordance with prior studies (Jacob and Neale, 1999; Clement and Tse, 2003 Salerno, 2014; Beuselinck et al., 2018). The SIZEMNC,LOSSMNC and HORIZONMNC all show a
negative relationship with the RFEMNC. The coefficient for the size of the MNC is -0.011, which indicates
that an increase in size of the MNC is related to a decrease in forecast accuracy. This is inconsistent with prior research (Lang and Lundholm, 1996), that proves that larger firms have higher forecast accuracy. The coefficient for loss of the MNC is -0.034. In accordance with prior research (Hwang et al., 1996), I find that firms reporting a loss have a lower forecast accuracy. The coefficient for the forecast horizon of the MNC is -0.185, which suggests that an increase in the forecast horizon is associated with a decrease in forecast accuracy. This is consistent with prior research (Barniv and Thomas, 2005; Ben, Choi and Kang, 2008; Luo, 2009 and Salerno, 2014) who discovered that shorter forecast horizons are associated with higher forecast accuracy. The results for COMPMNC, EARNINGS_SURPRISEMNC and NAFFMNCshow no significant relationship with forecast accuracy. This
is not consistent with prior research, (Lang and Lundholm, 1996; Barniv and Thomas, 2005; Ben et al., 2008; Luo, 2009, Salerno, 2014; Hope, 2003) who obtained evidence for significant negative relationships.
27
The second model tests hypothesis 2, which states if the relationship investigated in model one is influenced by the rule of law of a country. The second model includes RULE_OF_LAWMNC and an
interaction term between RULE_OF_LAWSUB and |DACC|SUB. The relationship between rule of law and
forecast accuracy is insignificant (-0.004). Furthermore, the interaction term is equally insignificant (0.174). The former coefficient indicates that the rule of law has no effect on the forecast accuracy of analysts. The latter coefficient indicates that the rule of law does not influence the relationship between subsidiaries FRQ and the forecast accuracy of analysts following the parent of the MNC. This provides evidence to reject hypothesis 2. The control variables for model two have similar coefficients as the control variables in model one
28 The +/- sign indicate the predicted positive (+) or negative (-) relationship between the dependent and the independent variables. The *,**,***, indicate the level of statistical significance at 10%, 5% and 1% levels for the relationships. Additionally, t-statistics are provided for all relationships, shown under the coefficients. The relative forecast error (RFEMNC)
is the measure for the forecast accuracy of the analyst about the parent of the MNC. The relative forecast error is calculated by dividing AFE by MAFE, whereby the ratio is subtracted by 1 and multiplied by -1. For calculation of the AFE and the MAFE, see the methodology section. The measure for the FRQ (|DACC|SUB), this is the aggregate measure for the absolute
value of the discretionary accruals of the subsidiary of the MNC. The aggregate measure is calculated by taking the sum of the absolute value of the discretionary accruals of all the MNCs subsidiaries divided by the number of subsidiaries for the specific MNC. The rule of law (RULE_OF_LAWSUB) is an aggregate measure of the rule of law, which is one of the WGI
indicators and provides the score of a country’s rule of law quality (Kaufman et al., 2009). The aggregate measure is calculated by taking the sum of the absolute value of the rule of law scores of all the MNCs subsidiaries divided by the number of subsidiaries for the specific MNC. ROAMNC is the absolute value of return on assets is calculated as the net income
before extraordinary items divided by the total assets of a firm. FREQMNC is the number of forecasts done by an analyst for
a specific MNC each year. The COMPMNC is the number of MNCs forecasted by an analyst. SIZEMNC is the natural logarithm
of the book value of the MNC’s total assets. LOSSMNC is a MNC which reports a loss in a specific year.
EARNINGS_SURPRISEMNC is earnings of a MNC that are above analyst’s expectations. HORIZONMNC is the forecast time
before a MNCs earnings announcement. The NAFFMNC is the number of analysts following a specific MNC each year. All
variables are winsorized at a 10% level.
Table 3: Regression Analysis
Independent Variables: Pred. Sign (1) (2)
Intercept 0.095*** 0.095*** |DACC|SUB - -0.217** -0.410 (-2.27) (-2.27) RULE_OF_LAWSUB - -0.004 (-0.53)
RULE_OF_LAWSUB X |DACC|SUB - 0.174
(0.53) ROAMNC + 0.340*** 0.340*** (7.31) (7.26) FREQMNC + 0.004*** 0.003*** (3.83) (3.84) COMPMNC - -0.001 -0.001 (-1.83) (-1.84) SIZEMNC + -0.011*** -0.012*** (-2.60) (-2.62) LOSSMNC - -0.034*** -0.034*** (-4.78) (-4.78) EARNINGS_SURPRISEMNC - 0.131 0.131 (2.62) (2.60) HORIZONMNC - -0.185*** -0.173*** (-15.09) (-15.09) NAFFMNC - 0.001 0.001 (0.23) (0.26)
INDUSTRY effects Included Included
YEAR effects Included Included
Adjusted R2 3.65% 3.64%
F-statistics 26.98*** 24.30***
29
Discussion and Conclusion
This research examined the relationship between subsidiary FRQ and the forecasts accuracy of the analysts following the parent MNC. The study used aggregate discretionary accruals to predict subsidiaries FRQ, the aggregate rule of law of a country as prediction for IQ and the relative forecast error as a prediction for the forecast accuracy. It was expected that higher subsidiaries FRQ influences the forecast accuracy of analysts following the parent MNC positively. The results show that lower FRQ at the subsidiary level is associated with lower forecast accuracy among analysts about the parent of the MNC. In accordance with this study’s expectations and prior research (Wilson and Wu, 2011; Salerno, 2014; Embong and Hosseini, 2018), this research shows that higher (lower) FRQ relates to a higher (lower) forecast accuracy among analysts.
Additionally, this research examined the influence of IQ of a subsidiary’s country on the relationship between subsidiaries’ FRQ and forecast accuracy. Prior literature (Beuselinck et al., 2018; Durnev et al., 2017) found that higher IQ is related to lower discretionary accruals. Beuselinck et al. 2018 found that managers of MNCs manage earnings more through subsidiaries in countries where regulation is weaker. Therefore, this study predicted that the IQ of a country influences the relationship between subsidiaries FRQ and forecast accuracy positively. However, the results show that the relationship is not influenced by the IQ of a subsidiaries country.
Previous studies (Wilson and Wu, 2011; Salerno, 2014) gave insight into how management accounting choices influence the forecasts of analysts. This report gives further insight in how managers accounting choices influence the forecast accuracy. Specifically, this study focused on managers of MNCs and their earning management practices at subsidiary level. Therefore, it might be useful for analysts of MNCs who would like to forecast the influence of management accounting choices on the forecasts they make about the parent of the MNC. Besides, this study might be useful to investors by determining the earnings management practices of managers of MNCs. The analysts’ forecasts are amongst managers most relevant benchmarks, which increases the pressure to meet or beat those (Graham et al., 2005). Investors may use this study to investigate EM practices of managers of MNCs, managers could use EM to meet or beat analyst forecasts.
The study contributes to the forecasting literature, by showing that FRQ of material subsidiaries matter for the forecast accuracy of analysts. Prior studies (Wilson and Wu, 2011; Salerno, 2014; Embong and Hosseini, 2018) found that FRQ of standalone companies is of influence on the forecast accuracy, this study shows that the subsidiaries FRQ also has influence on the forecast accuracy of analysts about the
30
parent MNC. The forecasts accuracy of analysts might be affected by the earnings management practices of managers of MNCs at subsidiary level. Moreover, this study contributes to the literature investigating the effect of IQ on FRQ. Prior studies (Beuselinck et al., 2018; Durnev et al., 2017) found that countries with weaker IQ have lower FRQ at subsidiary level. The IQ in a subsidiary’s country may therefore also influence the relationship between subsidiaries’ FRQ and forecast accuracy. This is the first study, to the best of my knowledge, that investigates the positive effect of IQ on FRQ of subsidiaries and therewith the forecast accuracy of analysts following the parent MNC.
The study has some drawbacks. First, the brokerage files are not available anymore in the I/B/E/S database, which prevented me from including broker specific control variables that have been used in prior studies regarding forecasts of analysts (Salerno, 2014; Ben et al., 2008; Barniv, 2005). The study does not include broker specific variables as brokerage size, brokerage industry, new analysts in the brokerage house and analysts who left de brokerage house. It might have affected the results, because prior research (Salerno, 2014; Ben et al., 2008; Barniv, 2005) found evidence that these brokerage specific variables are related to forecast accuracy. I could not control for these variables which might have affected my results.
Another limitation relates to the sample period analyzed. This is only a period of three years because the data of the material subsidiaries had to be hand-collected. Due to time limitation, it was not possible to use a larger sample for the study. Finally, the measure for IQ might not be captured, because the aggregate measure of rule of law that has been used is only focused on private companies from Europe. The rule of law variation is lower compared to a sample which also includes countries from outside Europe (for example Beuselinck et al., 2018).
Future research could further examine the relationship between subsidiaries FRQ and forecasts of analysts by extending the sample period and by using other forecast measures to analyze the effect on forecasts of analysts. For example, the effect on forecast dispersion and forecast revisions could be investigated. Additionally, further research could use a sample with subsidiaries from outside Europe. This study is only focused on European material subsidiaries, while the IQ differences in Europe are relatively small compared to a sample that consists of countries IQ from more continents. Therefore, further research might use a sample with subsidiaries from outside Europe, to better examine the effect of IQ on the relationship between FRQ of subsidiaries and forecast accuracy of analysts about the MNC.
31
References
Abarbanell, J., & Lehavy, R. (2003). Can stock recommendations predict earnings management and analysts’ earnings forecast errors?. Journal of Accounting Research, 41(1), 1-31.
Ahmed, A. S. and Duellman, S. (2011). “Evidence on the role of accounting conservatism in monitoring managers’ investment decisions”. Accounting and Finance, 51 (3), 6090-633.
Al-Dmour, A., Abbod, M., & Al-Balqa, N. (2018). The Impact of the Quality of Financial Reporting on Non-Financial Business Performance and the Role of Organizations Demographic’ Attributes.
Academy of Accounting and Financial Studies Journal, 22(1), 1-18.
Alonso, J. A., & Garcimartín, C. (2013). The determinants of institutional quality. More on the debate. Journal of International Development, 25(2), 206-226.
Ashbaugh, H. (2001). Non-US firms' accounting standards choice. Journal of Accounting and Public
Policy 20(2), 129−153
Athanasakou, V. E., Strong, N. C., Walker, M. (2009). Earnings management or forecast guidance to meet analyst expectations?. Accounting and business research, 39(1), 3-35.
Athanassakos, G., Kalimipalli, M. (2003). Analyst forecast dispersion and future stock return volatility.
Quarterly Journal of Business and Economics, 42(1-2), 57-78.
Ball, R., Kothari, S. P., & Robin, A. (2000). The effect of international institutional factors on properties of accounting earnings. Journal of Accounting and Economics, 29(1), 1-51.
Barker, R., & Imam, S. (2008). Analysts’ perceptions of ‘earnings quality’. Accounting and Business
Research, 38(4), 313-329.
Barniv, R., Myring, M. J., Thomas, W. B. (2005). The association between the legal and financial reporting environments and forecast performance of individual analysts. Contemporary Accounting
Research, 22(4), 727-758.
Barron, O. E., Kim, O., Lim, S. C., Stevens, D. E. (1998). Using analysts' forecasts to measure properties of analysts' information environment. Accounting Review, 73(4), 421-433.
Barron, O. E., Stanford, M. H., Yu, Y. (2009). Further evidence on the relation between analysts' forecast dispersion and stock returns. Contemporary Accounting Research, 26(2), 329-357.
Bartov, E., Givoly, D., & Hayn, C. (2002). The rewards to meeting or beating earnings expectations. Journal of accounting and economics, 33(2), 173-204.
Behn, B. K, Choi, J., Kang. (2008) Audit Quality and Properties of Analyst Earnings Forecasts. The
Accounting Review, 83(2), 327-349.
32
Bergstresser, D., & Philippon, T. (2006). CEO Incentives and Earnings Management. Journal of
Financial Economics, 80(3), 511-529.
Beuselinck, C., Cascino, S., Deloof, M., & Vanstraelen, A. (2018). Earnings management within multinational corporations. The Accounting Review, Forthcoming.
Bodnar, G. M., & Weintrop, J. (1997). The valuation of the foreign income of US multinational firms: A growth opportunities perspective. Journal of Accounting and Economics, 24(1), 69-97.
Bonacchi, M., Cipollini, F., & Zarowin, P. (2018). Parents’ use of subsidiaries to “push down” earnings management: Evidence from Italy. Contemporary Accounting Research, 35(3), 1332-1362.
Brown, L. D. (2001). How important is past analyst forecast accuracy?. Financial Analysts Journal,
56(6), 44-49.
Burgstahler, D. C., Hail, L., & Leuz, C. (2006). The importance of reporting incentives: Earnings management in European private and public firms. The accounting review, 81(5), 983-1016. Bushman, R., & Piotroski, J. (2006). Financial reporting incentives for conservative accounting: The
influence of legal and political institutions. Journal of Accounting and Economics, 42(1-2), 107-148. Bushman, R . M. & Smith, A. J. (2001). “Financial accounting information and corporate governance”.
Journal of Accounting and Economics, 32(2), 237-333
Chen, F., Hope, O. K., Li, Q., & Wang, X. (2011). Financial reporting quality and investment efficiency of private firms in emerging markets. The Accounting Review, 86(4), 1255-1288.
Cheng, P., Aerts, W., & Jorissen, A. (2010). Earnings management, asset restructuring, and the threat of exchange delisting in an earnings‐based regulatory regime. Corporate Governance: An International
Review, 18(5), 438-456.
Chin, C. L., Chen, Y. J., & Hsieh, T. J. (2009). International diversification, ownership structure, legal origin, and earnings management: Evidence from Taiwan. Journal of Accounting, Auditing &
Finance, 24(2), 233-262.
Choi, T. H., & Pae, J. (2011). Business ethics and financial reporting quality: Evidence from Korea. Journal of Business Ethics, 103(3), 403-427.
Christophe, S. E. (2002). The value of US MNC earnings changes from foreign and domestic operations. The Journal of Business, 75(1), 67-93.
Clement, M. B., & Tse, S. Y. (2003). Do investors respond to analysts' forecast revisions as if forecast accuracy is all that matters?. The Accounting Review, 78(1), 227-249.
DeAngelo, H., DeAngelo, L., & Skinner, D. J. (1994). Accounting choice in troubled companies. Journal
of accounting and economics, 17(1-2), 113-143.
Dechow, P. M., & Dichev, I. D. (2002). The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 77(1), 35-59.
